You can better meet your business objectives by architecting your data platform on the Snowflake Data Cloud with clear differentiation between corporate and customer systems.
Two of phData’s core principles are “iterate till done” and “build to operate.” At first, it may seem these two principles are at odds with each other. But even with large scale data platforms, these principles can holdfast. Using Snowflake along with a solid operational foundation can allow for quick iteration without sacrificing operational excellence.
Let’s explore how.
Establishing the Foundation
Building a data platform usually begins with an MVP.
An MVP establishes an initial architecture, governance, and operating model without having to build all the capabilities necessary for a customer system. The goal is to prove the first use-case and demonstrate real business value:
In the example above, the ingestion, data warehouse, and consumption tools and systems are categorized as corporate, with the understanding that nothing can impact customer systems.
Common outcomes include:
- Representative dashboards
- BI environment for analysts to explore data and build additional use cases
- Establishing a platform for data scientists to begin experimenting with the data to build business models.
The First Customer Data Product
Completing the first use-case and establishing a corporate data platform for BI and machine learning is a powerful capability for any company. However, inevitably a data product will be leveraged in the core day-to-day operations of the business.
In order to deliver on the required SLAs while keeping costs in check, the customer operating components will need to be instantiated on the data platform. This includes:
- Customer Systems for Ingestion
- Customer Ingestion Data Warehouse
- Customer Consumption Data Warehouse
- Customer Consumption Systems
Each of these will be sized and configured optimally for the customer data product. Additionally, the customer systems will have elevated release, change, and incident management processes.
The data model established in the customer systems can be leveraged by the corporate systems in order to not duplicate ingestion efforts. As data sets are established, the customer system SLAs will take precedence. The information architecture should be structured to account for this setup.
Additionally, the Snowflake Data Marketplace might be a revenue source. Establishing both data quality and operational excellence should be front and center as you serve your customers.
Continuing to Scale
As you identify and develop more customer data products, you should have a clear process for promoting them to the proper systems and processes. This takes a clear articulation of the platform strategy and its architecture. The onboarding process should make it clear which types of work.
If you run into any bumps or have any questions regarding the implementation of customer-facing data products on Snowflake, we’d love to help!
